The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions facilitate the ingestion of untrusted mathematical expressions which are then executed as shell command arguments.
Ingestion points: Mathematical expressions are passed as string arguments to scripts/sympy_compute.py and scripts/z3_solve.py within the Tool Commands section of SKILL.md.
Boundary markers: No delimiters or explicit instructions to ignore embedded control characters are provided, potentially allowing an attacker to inject shell commands via the expression strings.
Capability inventory: The skill utilizes the Bash tool to run commands via uv run, providing a mechanism for command execution.
Sanitization: There is no evidence of input validation or sanitization to ensure the strings passed to the Python scripts are strictly mathematical and do not contain malicious shell sequences.

lebesgue-measure